Pareto Optimized Adaptive Learning with Transposed Convolution for Image Fusion Alzheimer's Disease Classification

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is a progressive neurological disorder impacting cognition and memory.
• Multimodal imaging is vital for monitoring AD progression by revealing brain changes.
• Medical image fusion combines data from different modalities for enhanced understanding.

Purpose of the Study:

• To explore Pareto optimized deep learning for integrating MRI and PET images in AD diagnosis.
• To evaluate the efficacy of VGG11, VGG16, and VGG19 architectures for multimodal image fusion.
• To assess the performance of image fusion techniques using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Main Methods:

• Utilized pre-existing Visual Geometry Group (VGG) architectures (VGG11, VGG16, VGG19).
• Applied morphological operations and image manipulation for MRI and PET data alignment.
• Incorporated transposed convolution layers and Pareto optimization for hyperparameter tuning and feature fusion.

Main Results:

• VGG19 demonstrated superior performance compared to VGG16 and VGG11 in image fusion.
• Evaluated performance using metrics like SSIM, PSNR, MSE, and Entropy on the ADNI dataset.
• VGG19 achieved higher average SSIM scores across different disease stages (CN, AD, MCI) for both MRI and PET modalities.

Conclusions:

• Pareto optimized deep learning, particularly VGG19, is effective for fusing MRI and PET images in Alzheimer's disease research.
• The proposed method enhances the integration of multimodal imaging data for improved diagnostic insights.
• This approach holds promise for advancing the monitoring and understanding of Alzheimer's disease progression.